Specific interaction between gangliotriaosylceramide (Gg3) and sialosyllactosylceramide (GM3) as a basis for specific cellular recognition between lymphoma and melanoma cells.

In view of the possible role of glycosphingolipids in defining the specificity of cell-cell interactions, the key molecules for recognition of cell surface glycosphingolipids have been studied. In addition to previously suggested recognition mechanisms involving endogenous lectins and glycosyltransferases, an alternative possibility, based on carbohydrate-carbohydrate (Lex-Lex) interaction, has been raised (Eggens, I., Fenderson, B., Toyokuni, T., Dean, B., Stroud, M., and Hakomori, S. (1989) J. Biol. Chem. 264, 9476-9484). We now report a highly specific interaction between gangliotriaosylceramide (Gg3, GalNAc beta 1----4Gal beta 1----4Glc beta 1----Cer) and sialosyllactosylceramide (GM3, NeuAc alpha 2----3Gal beta 1----4Glc beta 1----Cer). The interaction requires a bivalent cation (Ca2+ or Mg2+) and can be inhibited by sialosyl 2----3 lactose, anti-GM3 antibody (DH2), anti-Gg3 antibody (2D4), or EDTA. The strength of interaction between GM3 liposome and the Gg3-coated plastic surface was highly density-dependent. The mouse lymphoma L5178 AA12 cell line (high expressor of Gg3) interacted specifically with the mouse B16 melanoma cell line (high expressor of GM3). The interaction was inhibited by 5 mM sialosyllactose, anti-GM3 antibody, anti-Gg3 antibody, and EDTA in analogy to GM3-Gg3 interaction. L5178 AV27, a genetically related variant clone which does not express Gg3, showed no interaction with B16 cells. Untreated AA12 cells, but not 2D4-treated AA12 cells or AV27 cells, interacted with GM3 coated on the plastic surface. These findings suggest a specific interaction between AA12 cells and B16 cells based on Gg3-GM3 interaction.

GM23100 and OIG CA42505 and by funds from the Biomembrane Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The abbreviations used are: GSL, glycosphingolipid; mAb, monoclonal antibody; DMEM, Dulbecco's modified Eagle's medium; FCS, fetal calf serum; Gg3, gangliotriaosylceramide; GM3, sialosyllactosylceramide; PG, paragloboside (nLc4); SPG, sialylparagloboside (IV3NeuAcnLc4); TBS, Tris-buffered saline; BSA, bovine serum albumin; GMl, I13NeuAc-GgOse4Cer. GSLs are abbreviated according to the recommendations by IUPAC-IUB Commission on Nomenclature (27), but the suffix OseCer is omitted. cell surface is as recognition sites for determining specificity of cell-cell interactions. This assumption is based on observations of dramatic changes in GSL composition in association with differentiation, development, and oncogenesis (1) and on inhibition of cell recognition occurring in morphogenesis and histogenesis by specific types of GSLS or their oligosaccharide sequences (2)(3)(4)(5)(6). The specific carbohydrate structures differentially expressed in each cell type and each stage of development are assumed to be recognized by complementary molecule(s) expressed at the surface of the counterpart interacting cell (5,6). The recognition molecules complementary to carbohydrates have been assumed to be proteins such as endogenous lectins and glycosyltransferases. Recently, an alternative possibility has been proposed that the molecule recognizing cell surface carbohydrate (Le") per se is carbohydrate (Le") (7), and the recognition system involves a homotypic interaction. In contrast, we now present evidence for specific interaction between two GSLs, G M~ and Gg3, and demonstrate clear heterotypic interaction between two cell types expressing GM3 and Gg3, respectively.
Liposomes labeled with [14C]cholesterol containing GSLs were prepared, and their interactions with GSL-coated plastic surface were studied as previously described (7). Briefly, 25 pg of dimyristoylphosphatidylcholine, 25 pg of cholesterol containing I4C-labeled material (25,000 cpm/pg), and 12.5 pg of the respective GSL were dissolved in 100 p1 of diethyl ether and injected slowly into 1.9 ml of prewarmed TBS with agitation in a Vortex mixer.
Diethyl ether was rotary evaporated, and liposome concentration was adjusted.
To prepare the solid phase polystyrene plastic surface coated with GSLs, the ethanol solution of each GSL (50 pg/ml) was appropriately diluted, placed in each well of a 96-well flat bottom plastic plate (Falcon, 0.3 ml/well volume; Becton-Dickinson, Oxnard, CA), dried a t 37 "C, and incubated with 1% BSA. To each GSL-coated well, 100 p1 (25,000 cpm) of liposome containing GSL and 14C-labeled cholesterol was added and incubated a t room temperature over a rotary shaker for 16 h in TBS with 1 mM CaC12 and 0.5 mM MgC1,. Each well was washed seven times with TBS with the aid of a very thin capillary tip (Eppendorf 200 Ultra Micro Tips, distributed through Brinkmann Instruments) linked to an aspirator. The remaining liposomes on the solid phase were extracted with isopropyl alcohol/ hexane/water (55:25:20), and radioactivity was counted. Interaction between liposomes containing various GSLs was studied utilizing the fact that the Gg3 liposome adheres strongly to the polystyrene surface and is not readily washed off (e.g. 14,000 k 1,200 cpm out of 25,000 cpm of liposomes added per well were bound). Thus, Gg, liposome strongly adherent on the polystyrene surface after incubation with 1% BSA a t room temperature for 2 h provided solid phase-affixed Gg3 liposome capable of interaction with other liposomes (see Fig.  1C). Freshly prepared liposomes were used in all experiments.
Mouse B16 melanoma cells (expressing a high level of GMj, defined by mAb DH2) (9) were purchased from American Type Culture Collection (Rockville, MD). Mouse T cell lymphoma line L5178 AA12 (expressing a high level of Gg,, defined by mAb 2D4) (10) was cloned from 1C1 (12) in this laboratory. A variant clone, L5178 AV27, which does not express Gga (12), was cloned originally in the laboratory of Christopher Henny (then a t Fred Hutchinson Cancer Re-20159 search Center, Seattle, WA). All cells were grown in RPMI 1640 containing 10% FCS.

RESULTS AND DISCUSSION
We observed specific interaction of the GM3 liposome with Ggn coated on solid phase, whereas liposomes containing SPG, PG, GM,, or Ggn showed no such interaction. The G~3-Gg3 interaction was proportional to the quantity of Ggn coated, and GMn liposome did not interact with Ggn coating pretreated with anti-Gg3 mAb 2D4 (Fig. lA). G M~ liposome interacted in a dose-dependent manner with Gg3-coated solid phase but not with PG-, SPG-, or GMl-coated solid phase (Fig. 1B). A repulsive interaction was observed between GM3 liposome and G~c o a t e d surface (Fig. 1B). GM3

liposome interacted with
Ggn liposome in a dose-dependent manner, whereas Gg3 liposome did not interact with Gg3 liposome per se (Fig. IC). The interaction of G M~ liposome with Gg3-coated solid phase varied with the respective densities of the two GSLs. The GMg density-dependent increase of G M~ liposome was clearly observed a t low Gg3 concentration on solid phase (Fig. ID). Thus Fig. 3. Open and shaded columns show adhesion of AA12 and AV27 cells, respectively. Effects of (respectively) 10 mM EDTA, 5 mM sialic acid, 5 mM sialyllactose, a n t i -G~~ antibody DH2, and anti-Gg3 mAb 2D4 are indicated on the ordinate. B16 melanoma cells grown in 24-well plates and L5178 lymphoma cells were treated, respectively, with DH2 solution (2 pg/ml) and 2D4 solution (5 pg/ ml), left for 30 min at room temperature, washed with serum-free fresh DMEM, and incubated with lymphoma cell suspension and melanoma cell layer for 5 h. Values represent means of triplicate determinations.  (9) were grown on 24-well plates (Falcon, Oxnard, CA) for 24 h in DMEM. The majority of adherent cells were washed with serum-free medium. AA12 cells (Gg, expressors) or AV27 cells (Gg3 nonexpressors) were grown in DMEM containing 2% FCS, 1 g/liter glucose, and 50 pCi of [3H]GlcNH2 for 24 h. The labeled AA12 (71,000 cpm/105 cells) and AV27 (33,000 cpm/105 cells) were then collected by centrifugation, suspended in DMEM containing 2% FCS, and the cell population adjusted to 6 X 106/ml. This suspension (1 ml) was added to each well on which B16 melanoma cells were grown as adherent culture for appropriate times. The incubated plates were then washed 2 times with serum-free fresh medium. L5178 lymphoma cells not adhering to B16 cells were washed out. The remaining cells in each plate were trypsinized, collected, and counted. 0, AA12 cells; 0, AV27 cells. The interaction of GM3 liposome with GgS coated on the plastic surface was abolished in the presence of 5 mM sialyllactose but not 1 mM sialyllactose, 5 mM sialic acid, or 5 mM lactose and was strongly enhanced in the presence of Ca2+ or Calculated by Scatchard plot using various amounts of liposomes containing several concentrations of GM3 (2.5-12.5 pg/ml) on a constant solid phase of Gg, (1 pg/well). This value can be regarded as an apparent KD value.

Mg2+ and abolished in the presence of EDTA (Table I, A).
The interaction of G M~ liposome with Gg3 liposome affixed to the plastic surface was also strongly promoted in the presence of Ca'+ or M$+ and inhibited by EDTA or 5 mM sialyllactose (Table I, B). Involvement of bivalent cations in carbohydratecarbohydrate interaction has been reported for Lex-Lex interaction (7); there is considerable experimental data showing the reaction of carbohydrates with various metal cations including Ca' ' and Mg2+ (cited in Ref. 7).
Mouse T cell lymphoma L5178 AA12 expressing high levels of Gg, but not the variant clone L5178 AV27 that does not express Gg, (12) interacted with mouse melanoma B16, which is a high expressor of G M~ (9) (Fig. 2). Time-dependent quantitative adhesion of AA12, but not of AV27 cells, on B16 melanoma is shown in Fig. 3A. AA12 cells specifically bound to &,-coated solid phase in a dose-dependent manner. This interaction was inhibited when AA12 cells were treated with anti-Gg3 mAb 2D4 or G~~-c o a t e d solid phase was pretreated with anti-(& mAb DH2 (Fig. 3B). The strong interaction between AA12 and B16 cells was inhibited by 5 mM sialyllactose but not by 5 mM sialic acid. The interaction was also inhibited by treatment of B16 cells with mAb DH2, treatment of AA12 cells with mAb 2D4, the absence of bivalent cation, or the presence of EDTA (Fig. 4). Both AA12 and AA27 clones were derived from L5178 cells and are very similar genotypically and phenotypically, except that AA12 expresses Gg3 whereas AA27 does not. We observed that AA12 (but not AA27) interacted with B16 melanoma in analogy to the Gg3-G M~ interaction, i.e. the interaction was inhibited by 5 mM sialyllactose and mAbs 2D4 and DH2. Thus, it can be assumed that the specific cellular recognition between AA12 and B16 is based on molecular interaction between Gg3 and GM3. This is supported by the fact that the AA12-B16 interaction was inhibited by treating AA12 cells with mAb 2D4 or treating B16 cells with mAb DH2; these antibodies are highly specific to their respective GSL antigens and do not cross-react with glycoproteins (11, 12).

Glycosphingolipid to Glycosphingolipid Interaction
Minimum energy conformation models of Gg3 and G M~ have been constructed based on hard sphere exoanomeric calculations (13, 14) and on glycosidic torsion angles (d/*) of GalNAc@l-liGal (54/10") and  for Ggn, and Galpl+4Glc (55/4") for GM3. The models show that one side of each molecule is hydrophobic while the other side is hydrophilic (Fig. 5, upper  part). A three-dimensional Covey-Pauling-Koltun model of the two molecules was constructed accordingly. We found that the hydrophobic sides but not the hydrophilic sides of Gg3 and GM3 fit closely together when the distal ends of the hydrophobic surfaces were placed vis-a-vis and the axes were oriented obliquely a t a 40-65" angle ( Fig. 5, lower part). The hydrophilic area adjacent to the hydrophobic surfaces (side 1) of GMn and Gg3 (Fig. 5, lower part) may stabilize hydrophobic adhesion between these two surfaces. Furthermore, bivalent cations may stabilize the interaction between the hydrophilic surfaces (side 2) of GM3 and Gg3 through chelation.  (13,14). One side (side 1) of each molecule exposes a larger hydrophobic area, as indicated by an abundance of hydrogen atoms (white balk) attached to carbon (black balk). The other side (side 2) is hydrophilic, as indicated by an abundance of oxygen (gray shaded balls). Hydrogen atoms linked to oxygen are omitted from this model. Ceramide portions are oriented perpendicularly toward the hydrophilic sides; thus the hydrophobic surfaces (side 1) are exposed on the outer cell membrane. The lower part illustrates the side view of the possible interaction between GgR and via their hydrophobic (side 1) surfaces. Lower left, CPK model; lower right, an outline of two molecules. GXq3 (dark shaded) on the membrane surface I is-in the front with its hydrophobic surface (side 1) facing behind. Ggn Although there has been extensive discussion of the role of carbohydrates in cell recognition, it has been viewed mainly in terms of interaction between carbohydrates and proteins, particularly endogenous lectins (15-17). However, expression of lectins at the cell surface in mammalian tissues is highly limited, being observed mainly in hepatocytes (18) and certain tumor cells (19). The majority of endogenous lectins may be located intracellularly, and their functions may not be ektobiological (20, 21). Glycosyltransferases, often discussed in this same context (22), are similarly limited to specific types of cell surface expression, e.g. in sperm cells (23).
Cell recognition and adhesion are continuous multistep processes initiated by a specific cell-cell recognition between either homotypic or heterotypic molecules. We have previously shown that Le"-Le" interaction could be the initial triggering mechanism for homotypic aggregation in mouse teratocarcinoma F9 cells (7). Cell adhesion, initiated by this mechanism, subsequently involves a cell type-nonspecific adhesion protein present pericellularly (e.g. cadherin, laminin, and fibronectin) and a corresponding receptor (24)(25)(26).
In analogy to this example, we have now demonstrated heterotypic cell adhesion in which the initial specificity is based on Gg3-GM3 interaction and followed by a step involving an unidentified adhesive protein and receptor, probably a member of the integrin family.
Cell type-specific glycosylation patterns and their continuous alteration during development, differentiation, and oncogenesis are well established (1). The present study provides additional evidence that carbohydrate-carbohydrate interactions may sometimes provide the initial cell type-specific recognition prior to cell type-nonspecific adhesion.